Form Recognizer: language detection into Layout model?

Keir 20

Current state as I have understood:

Read model: document as input, ocr exists, language detection exists (multiple languages returned)

Layout model: document as input, ocr exists, table detection exists, no language detection

Text analytics: text as input, output 1 single language

End goal: to get table detected & most popular languages detected via one API call (otherwise time-consuming etc)

Is there possibility to combine models together? (e.g somehow include language detection into Layout model like Read model has)
Is there possibility to detect (multiple) languages by text?

VasaviLankipalle-MSFT 15,956 Reputation points

2023-03-08T01:39:35.2566667+00:00

Hi @Keir , Thanks for using Microsoft Q&A Platform.

That's true layout model does not support language detection.

There is no such way to combine models, one option is custom composed model. Since your use case is different, I don't think it will help in language detection.

Could you please share with us the specific scenario with which you want language detection in the layout model?

Regards,
Vasavi
VasaviLankipalle-MSFT 15,956 Reputation points

2023-03-09T02:26:36.8433333+00:00

@Keir , did you get a chance to check my response?
Keir 20 Reputation points

2023-03-09T19:54:57.3433333+00:00
Yes.

I have such use case when I'd like to get following info from document:

OCR itself (words, lines or just characters with coordinates)

1-3 popular languages detected

table detection

And then I can work with data to put our business context and make decisions.

It is not reasonable at the moment to make 2 separate queries (Read by doc + Layout by doc) because time-consuming & 2 times money. Also recommended Custom model seems to be much more expensive.
I could handle to make 2 queries (Layout by doc + Language detection by text) but Text Analytics returns only 1 language. But our business documents are having also 2 languages on it and I need to know it.

Actually I have working flow with one competitor in LIVE but I wanted to switch to Azure Cognitive services and it is a bit blocker. Do you have in near future plans to support something like that?
VasaviLankipalle-MSFT 15,956 Reputation points

2023-03-10T04:55:56.3933333+00:00

Hi @Keir , thank you for sharing this with us. Please allow some time to check internally and get back to you.
VasaviLankipalle-MSFT 15,956 Reputation points

2023-03-14T19:24:53.9566667+00:00

Hi @Keir , we received your feedback that the answer provided on the thread was not helpful.

Kindly let us know what we could have done better to improve the answer and make your engagement experience good. We are here to help you and strive to make your experience better and greatly value your feedback.

As mentioned earlier, if you could send an email to us at azcommunity@microsoft.com regarding your scenario, we can happily set up a call with the PG team.

If you wish, you may re-surveying/rating for the engagement you received on the thread. Your feedback is very important to us.

Looking forward to you reply. Much appreciate your feedback!

Regards,
Vasavi

Accepted answer

VasaviLankipalle-MSFT 15,956 Reputation points

2023-03-13T20:12:28.12+00:00

Hi @Keir , thank you for taking the time in sharing the use case with us.

As layout model does not support language detection. There is no such way to combine models, one option is custom composed model. Since your use case is different towards language detection.

The product team is interested in your scenario. PG wanted to understand more about it in order to provide the best feature. If possible, could you please send us an email to azcommunity@microsoft.com with the below details so that we can set up a call between you and PG to discuss more on this?

Subject: Attn: Vasavi
Thread URL: Link to this thread.
Use case/Scenario:

I hope this helps.

Regards,
Vasavi

-Please kindly accept the answer if you feel helpful to support the community, thanks.
Please sign in to rate this answer.

0 comments No comments
Sign in to comment

Share via

Form Recognizer: language detection into Layout model?

0 additional answers